Violence exposure and social deprivation is associated with cortisol reactivity in urban adolescents

Psychoneuroendocrinology 111 (2020) 104426

Contents lists available at ScienceDirect

Psychoneuroendocrinology journal homepage: www.elsevier.com/locate/psyneuen

Violence exposure and social deprivation is associated with cortisol reactivity in urban adolescents

T

Melissa K. Peckinsa, Andrea G. Robertsa, Tyler C. Heina,b, Luke W. Hydea,c,d, Colter Mitchelld, Jeanne Brooks-Gunne, Sara S. McLanahanf, Christopher S. Monka,c,d, Nestor L. Lopez-Durana,⁎ a

Department of Psychology, University of Michigan, United States Serious Mental Illness Treatment Resource and Evaluation Center (SMITREC), Office of Mental Health and Suicide Prevention, Department of Veterans Affairs, United States c Center for Human Growth and Development, University of Michigan, United States d Institute for Social Research, University of Michigan, United States e Teachers College, Columbia University, United States f Department of Sociology, Princeton University, United States b

A R T I C LE I N FO

A B S T R A C T

Keywords: Cortisol reactivity Violence Social deprivation Adolescence

The present study tested how two different dimensions of childhood adversity, violence exposure and social deprivation, were associated with the cortisol response to the Socially Evaluated Cold-Pressor task in a sample of 222 adolescents (n = 117 girls, n = 167 African American). Participants were part of the Fragile Families and Child Wellbeing Study, a probability sample of births in large US cities (> 200,000) between 1998 and 2000. Our subsample includes births in three cities: Detroit, Toledo, and Chicago. The study design called for an oversampling of births to unmarried parents (3:1) which led to a large number of minority and economically disadvantaged adolescents. When children were ages 3, 5, and 9, mothers reported on exposures to violence and social deprivation that occurred in the past year. Exposures from the three waves were averaged to reflect violence exposure and social deprivation during childhood. Greater levels of violence exposure from ages 3 to 9 were associated with a blunted cortisol response to stress at age 15, even after controlling for social deprivation and other factors known to influence cortisol reactivity. Social deprivation from ages 3 to 9 was not associated with the cortisol response to stress; though in an exploratory analysis, social deprivation moderated the association between violence exposure and cortisol peak activation. In line with the Dimensional Model of Adversity and Psychopathology, these findings suggest that experiences of violence, but not social deprivation, during childhood may contribute to cortisol blunting that has been previously reported in samples with high levels of social deprivation. Findings from the present longitudinal study on a relatively large sample of under-represented minority youth provide insight into the ways two different dimensions of childhood adversity impact the cortisol response to stress.

1. Introduction Dysregulation of the hypothalamic-pituitary-adrenal (HPA) axis is a risk factor for many psychological and physical health problems across the lifespan (Agorastos et al., 2018). Adverse experiences during childhood can lead to atypical HPA-axis functioning (Koss and Gunnar, 2018). Most studies, however, do not distinguish between different types of adversity (e.g., threat versus low resources), but instead combine all adverse experiences into one index, making it difficult to disentangle how different forms of adversity impact functioning of the HPA-axis (McLaughlin et al., 2014). Examining different types of

⁎

childhood adversity as qualitatively different dimensions may help explain inconsistencies in the HPA-axis reactivity literature (Busso et al., 2017; Kuhlman et al., 2015). Moreover, most studies that examine the impact of adversity on HPA-axis reactivity are not based on representative samples and do not include large numbers of AfricanAmerican youth (e.g., Busso et al., 2017). Instead, they focus on narrowly characterized trauma or samples exposed to violence (e.g., trauma survivors, orphans). In the present study we examined how two dimensions of childhood adversity – violence exposure and social deprivation – relate to reactivity of the HPA-axis in a diverse sample of adolescents with high rates of socioeconomic disadvantage (Coulton

Corresponding author at: 1004 East Hall, 530 Church Street, Ann Arbor, MI 48109, United States. E-mail address: [email protected] (N.L. Lopez-Duran).

https://doi.org/10.1016/j.psyneuen.2019.104426 Received 2 April 2019; Received in revised form 2 August 2019; Accepted 2 September 2019 0306-4530/ © 2019 Published by Elsevier Ltd.

Psychoneuroendocrinology 111 (2020) 104426

M.K. Peckins, et al.

longitudinal, prospective reports of violence exposure and social forms of deprivation during childhood (i.e., child neglect, low social support, low neighborhood social cohesion), using two-piece multilevel growth curve modeling with landmark registration. Previous research has shown that experiences high in threat may affect activation of the HPAaxis, whereas experiences high in social deprivation may be associated with recovery of the HPA-axis (Kuhlman et al., 2015). Therefore, we expected greater experiences of violence would be associated with blunted cortisol reactivity and peak activation, whereas greater experiences of social deprivation would be associated with blunted cortisol recovery (i.e., a less steep decline in cortisol). Given that past research found an association between hypo-reactivity of the HPA-axis and interpersonal violence exposure, but not poverty (Busso et al., 2017), we hypothesized exposure to violence would be associated with a blunted cortisol response over and above the effects of social deprivation. As previous research has also identified low levels of social support as a risk factor for blunting of the HPA-axis following violence exposure (Aiyer et al., 2014), we performed an exploratory analysis to test whether the association between violence exposure and the cortisol response to stress was conditional on level of social deprivation.

et al., 2007). The Dimensional Model of Adversity and Psychopathology (DMAP; McLaughlin et al., 2014) suggests that different types of adverse experiences can uniquely impact neurobiological functioning. For example, experiences high in threat such as violence exposure and victimization (i.e., “violence exposure”) are expected to influence the development of cortico-limbic circuitry involved in fear learning and salience processing, thereby impacting HPA-axis reactivity (Busso et al., 2017). In contrast, experiences high in social and environmental deprivation, including neglect or living in poverty, are hypothesized to have a more global impact on cortical thickness and white matter integrity, and to not have the same impact on the HPA-axis (Busso et al., 2017; McLaughlin et al., 2017). However, despite DMAP being a prominent theory in the literature, it has not been widely tested in studies of the HPA-axis. The few studies that have tested the DMAP in relation to HPA-axis functioning have consisted of samples with low levels of deprivation or have measured deprivation as a single indicator of poverty (e.g., income to needs ratio). Experiences high in threat such as physical abuse and interpersonal violence during childhood have been associated with both hyper- and hypo-reactivity of the HPA-axis (Busso et al., 2017; Kuhlman et al., 2015), and this discrepancy may be explained by the developmental period during which HPA-axis function was measured (Koss and Gunnar, 2018; Roberts and Lopez-Duran, 2019). For example, one longitudinal study found that the developmental time course of HPAaxis function differs for sexually abused and comparison girls (Trickett et al., 2010). Relative to comparison girls, sexually abused girls started out with elevated cortisol levels in childhood and transitioned to blunted cortisol levels during adolescence and into adulthood, years after the maltreatment experiences occurred (Trickett et al., 2010). Bosch et al. (2012) also marked adolescence, particularly the onset of puberty, as an important transition point for HPA-axis function following childhood adversity, but in the opposite direction. Greater exposure to adversity between the ages of 6 and 11 years was associated with higher cortisol levels during adolescence (Bosch et al., 2012). However, this study included experiences of both threat and deprivation in their measure of adversity (Bosch et al., 2012). Although there is evidence that childhood experiences influence HPA-axis reactivity during adolescence, it is unknown whether experiences of threat, deprivation, or both contributed to dysregulation of the HPA-axis. Similar to research on threat, high social and environmental deprivation has been linked to hyper- and hypo-reactivity of the HPA-axis, or, in some cases, has been found to not be associated with HPA-axis reactivity (Barrington et al., 2014; Busso et al., 2017; Kapuku et al., 2002; Koss and Gunnar, 2018; Wismer Fries et al., 2008). Busso et al. (2017) have suggested that discrepancies in the deprivation literature may be due to co-occurring experiences of threat (e.g., violence exposure) that are not being accounted for in statistical models. Contrary to previous studies of poverty and the HPA-axis (Koss and Gunnar, 2018), Busso et al. (2017) found that poverty was not associated with HPA-axis reactivity on its own or when controlling for interpersonal violence exposure. However, Busso et al. (2017) included a single measure of poverty (i.e., income to needs ratio) that may not be a complete representation of youths’ experiences of deprivation. Furthermore, Busso et al. (2017) only measured HPA-axis reactivity and not recovery. Research suggests different forms of adversity may uniquely impact each phase of the HPA-axis response to stress (Kuhlman et al., 2015). Thus, an important contribution to this literature will be to examine how experiences of threat, deprivation, and the interplay between the two impact HPA-axis reactivity, peak activation, and recovery (Koss and Gunnar, 2018; McLaughlin et al., 2014). The present study examines how experiences high in threat and deprivation affect each phase of the cortisol response to a laboratory stress paradigm in a community sample of adolescents who were at increased risk for experiencing childhood adversity due to contextual characteristics (e.g., high levels of poverty). We tested these aims with

2. Methods 2.1. Participants Participants included a subsample of 237 adolescents (n = 124 girls) from the Study of Adolescent Neural Development (SAND; Goetschius et al., 2019; Hein et al., 2018) living in Detroit, Toledo, and Chicago who participated in the Fragile Families and Child Wellbeing Study (FFCWS; Reichman et al., 2001). The FFCWS is a multi-stage, stratified, probability sample of births in large US cities (> 200,000) between 1998 and 2000 (Reichman et al., 2001). The study design called for a 3:1 oversample of non-marital births, which led to a large number of births to low income, minority parents (Reichman et al., 2001). Data were collected on the mother, father, and child at birth and again when the child was approximately 1, 3, 5, 9, and 15 years old. The current study is based on births in three sample cities: Detroit, Toledo, and Chicago. There were 428 families in the original Detroit and Toledo subsamples of FFCWS; we attempted to contact all families from these two sites. To increase the number of participants, we attempted to contact 78 families from the Chicago subsample. Of the 506 FFCWS families contacted, 237 families participated in SAND data collection. Families that agreed to participate in SAND data collection did not differ from families that refused or were unreachable on annual household income at the six waves of the FFCWS (ps = .11–.84). However, nonparticipation was associated with mother reports of community violence exposure when adolescents were ages 3, 5, and 9 (χ2 = 6.72, df = 1, p < .05, V = .12). Sixty-three percent of all mothers from families that participated in SAND endorsed at least one form of community violence when adolescents were ages 3, 5, or 9 compared to 51.10% of families that did not participate. Within the Detroit, Toledo, and Chicago subsamples, mothers’ self-report of race/ ethnicity at the time of the child’s birth did not differ between mothers who did and did not participate in SAND (ps = .13–.49). However, mothers’ self-report of race/ethnicity at the time of the child’s birth differed between the SAND sample and the larger FFCWS (χ2 = 86.32, df = 3, p < .01, V = .13). The majority of mothers who participated in SAND data collection identified as Black/African American and nonHispanic (75%), and also included mothers identifying as White/Caucasian and non-Hispanic (16%), Hispanic (6%), and other (3%). In the FFCWS, 46% of mothers identified as Black/African American and nonHispanic, 21% identified as White/Caucasian and non-Hispanic, 29% identified as Hispanic, and 4% as other. At age 15, participants in the SAND sample visited the University of Michigan where, following assent and parental consent, they completed questionnaires, provided biological samples, and completed a 2

Psychoneuroendocrinology 111 (2020) 104426

M.K. Peckins, et al.

Following the SECPT, participants were accompanied to a different room that included a comfortable couch and a large flat-screen television where they watched a neutral non-emotionally inducing video alone for a total of 60 min (n = 164) or completed non-stressful questionnaires for 60 min (n = 64). The videos included three educational documentaries by National Geographic (i.e., The Appalachian Trail, Ballad of the Irish Horse, Ocean Drifters) that had no violent or distressing content. Participants were redirected if they engaged in a different activity (e.g., playing video games with a cell phone) during this period (Kuhlman et al., 2015). The post-stressor activity was not recorded for the remaining subjects (n = 9).

laboratory stress paradigm. Only a subset of the collected measures were used in the present study. All procedures were approved by the University of Michigan Institutional Review Board. To limit selection effects, we provided families with transportation to and from the University of Michigan and when transportation was declined, compensated families for mileage and gas. We held visits on weekends to accommodate work and school schedules and compensated families for their time spent at the University of Michigan. Families were allowed to bring siblings and other family members to the University of Michigan, and we provided child care and lunch for the entire family. Participants were excluded from the present study if they refused to provide saliva samples (n = 10) or if they had elevated cortisol concentrations (greater than 3 SD from the Mean) due to assay interference from improper saliva collection or food/drink consumption that produced a false result (n = 5). The final sample resulted in 222 adolescents (n = 117 girls) who identified as Black or African American/nonHispanic (n = 167), Caucasian/non-Hispanic (n = 29), Hispanic (n = 9), biracial (n = 14), and other (n = 3). At the time of the adolescent’s birth, 34% of birth mothers in this sample reported having less than a HS diploma or not completing their GED. From birth through age 15, 40–46% of families were within 100% of the poverty threshold, and 20–28% of families were between 100–200% of the poverty threshold (based on the year data were collected and the number of adults and children in the household). Thus, adolescents in this sample experienced high rates of socioeconomic hardship throughout childhood and adolescence. Adolescents in this sample were also exposed to greater rates of community violence than other samples from the FFCWS (e.g., James et al., 2018). When adolescents were ages 3, 5, and 9, 32–43% of mothers witnessed someone get beat up in the community, 15–19% witnessed someone get attacked with a weapon in the community, and 7–12% witnessed someone get shot at in the community.

2.2.2. Salivary cortisol Adolescents provided a total of 9 saliva samples via passive drool throughout the SECPT protocol: immediately prior to the onset of the SECPT (Sample 1) and at 3 min, 15 min, 25 min, 30 min, 35 min, 40 min, 45 min, and 60 min post-SECPT onset (Samples 2–9). We collected samples at these times because they maximize the ability to capture the cortisol stress response, since peak levels of cortisol in saliva are observed approximately 25 min after beginning the SECPT and cortisol levels return to baseline approximately 60 min post-SECPT onset (Schwabe and Schächinger, 2018). Saliva samples were stored in a -20c freezer until time of assay. Saliva samples were assayed for cortisol by the University of Michigan Core Assay Facility with a high sensitivity enzyme immunoassay kit (Salimetrics, State College, PA). The assay sensitivity was < 0.007 μg/dL and assay range was 0.012–3.000 μg/dL. The inter-assay and intra-assay coefficients of variance were 9.72% and 6.25%, respectively. A total of four cortisol samples were missing due to insufficient saliva volume. 2.2.3. Violence exposure and social deprivation Violence exposure and social deprivation were constructed using data from the FFCWS when adolescents were ages 3, 5, and 9 years old. Our violence exposure and social deprivation constructs were each designed to reflect childhood exposures that were directed: (1) at the child, (2) exposures in the home between caregivers, and (3) exposures within the community (Hein, 2019). Violence exposure was conceptualized as: (1) violence directed at the child by a primary caregiver (physical and emotional abuse), (2) intimate partner violence exposure, and (3) exposure to community violence. Social deprivation was conceptualized as: (1) physical and emotional neglect directed at the child by a primary caregiver, (2) the absence of romantic partner support for the primary caregiver, and (3) lack of social cohesion within the community (Hein, 2019). At ages 3, 5, and 9, mothers reported the frequency at which physical (5 items) and emotional abuse (5 items) occurred in the past year on a 7-point Likert scale ranging from “never happened” to “more than 20 times” using the Parent-Child Conflict Tactics Scale (CTS-PC; Straus et al., 1998). The CTS-PC is a widely used parent report measure of conflict and violence in intimate relationships. The FFCWS eliminated 8 questions from the CTS-PC that asked about severe physical maltreatment. In the present study, physical abuse items included “hit him/her on the bottom with something like a belt, hairbrush, or stick, or some other hard object”, and “shook him/her.” Emotional abuse items included whether the parent had “sworn or cursed at” or “called him/her dumb or lazy or some other name like that.” Items were summed to calculate total physical and emotional abuse experienced at each age (Cronbach’s α = .77–.81). Nearly all mothers endorsed at least one instance of physical or emotional abuse at ages 3, 5, or 9 (97.89%). Mothers reported any physical (2 items), emotional (3 items), or sexual (1 item) intimate partner violence inflicted by the child’s biological father or a current romantic partner (of the mother) on a 3-point Likert scale ranging from “never” to “often” (Hunt et al., 2017). Physical intimate partner violence items included “he slapped or kicked you” and “he hit you with his fist or a dangerous object.” Emotional intimate partner violence items included “he tried to isolate you from family and

2.2. Measures 2.2.1. Socially Evaluated Cold-Pressor Task (SECPT) The SECPT is a laboratory stress paradigm that contains both physiological and social-evaluative components (Schwabe et al., 2008). During the SECPT, subjects immerse their hand in ice water while being watched by a research assistant and videotaped (Schwabe et al., 2008). Using samples of adults, previous studies have demonstrated that the SECPT leads to an increase in cortisol between 34–127% above baseline levels, with 48–84% of adults increasing in cortisol by at least 15.5% (Schwabe and Schächinger, 2018). The majority of adults reached their peak cortisol concentration at 25 min post-SECPT onset and returned to baseline levels of cortisol approximately 60 min post-SECPT onset (Schwabe and Schächinger, 2018). In the present study, trained research assistants filled a bucket with ice water until approximately 1/4th full and stirred the water until the temperature was between 1–3 °Celsius. Participants were brought into the experimental room and told that they would be videotaped and the videos would be analyzed for facial expressions. Participants were then asked to put their non-dominant hand inside a bucket of cold water kept at 1–3 °Celsius and were instructed to keep their hand inside the water for as long as possible while looking at the video camera placed on a tripod in front of the participants. If participants removed their hand before 60 s, they were instructed to place their hand back in the water for a second trial. This procedure was repeated until at least 60 s of hand immersion was achieved or 3 min passed. Anyone who had their hand in the water for 3 min was asked to remove their hand from the water. All participants were instructed to look at the camera for a total of 3 min regardless of total hand immersion time. Consistent with the current recommendations for administering the SECPT (Schwabe and Schächinger, 2018), all participants spent a minimum of 45 min at the lab prior to completing the SECPT. To avoid time of day effects, all participants completed the SECPT in the afternoon (M = 14:17 h, SD =39.96 min, Range = 12:56-16:07). 3

Psychoneuroendocrinology 111 (2020) 104426

M.K. Peckins, et al.

friends” and “he tried to prevent you from going to work and/or school.” The sexual intimate partner violence item was “he tried to make you have sex or do sexual things you didn’t want to do.” Items were summed at each age, and if the child did not live with the mother at least half of the time for a given age, intimate partner violence for that wave was coded as missing (Cronbach’s α = .37–.60). Only 19.83% of mothers reported at least one instance of intimate partner violence at ages 3, 5, or 9. Mothers also reported on community violence exposure within the past year on a 5-point Likert scale ranging from “never” to “more than 10 times” (Zhang and Anderson, 2010). At ages 3 and 5, items on witnessing or being the victim of beatings, attacks with a weapon, shootings, and killings (witness only) were summed (Cronbach’s α = .85). At age 9, items on witnessing beatings, attacks with a weapon, and shootings were summed (Cronbach’s α = .73). Over half of all mothers reported at least one instance of community violence exposure at ages 3, 5, or 9 (62.87%). Mothers reported the frequency at which physical (4 items) and emotional neglect (1 item) occurred in the past year using the CTS-PC (Straus et al., 1998) on the same scale as the physical and emotional abuse items. Physical neglect items included asking whether the parent was ever “not able to make sure your child got the food he/she needed” and “so drunk or high that you had a problem taking care of your child.” The emotional neglect item asked whether the parent had been so caught up with their own problems that they were not able to show love to the child. Items were summed to calculate total physical and emotional neglect experienced at each age (Cronbach’s α = .15–.60). Only 32.91% of mothers reported at least one instance of neglect at ages 3, 5, or 9. Mothers reported on the romantic partner support between themselves and the target child’s birth father or current romantic partner with six items on a 3-point Likert scale ranging from “never” to “often” (Manuel et al., 2012). Romantic partner support included “how frequently (the current romantic partner) expresses love and affection (for the mother).” Items were summed and reverse coded (higher scores indicate less romantic support; Cronbach’s α = .51–.78). If the target child did not live with the mother at least half of the time for a given age, romantic partner support for that age was coded as missing. Over half of all mothers in the SAND sample experienced at least one instance of low social support at ages 3, 5, or 9 (69.20%). Mothers were also asked to complete four questions on neighborhood social cohesion (Morenoff et al., 2001). Items included asking the primary caregiver whether they felt “this is a close-knit neighborhood.” Items were scored on a 5-point Likert scale ranging from “strongly agree” to “strongly disagree” and were summed at each wave with higher scores indicating less social cohesion (Cronbach’s α = .63–.79). The majority of mothers endorsed at least once instance of low neighborhood social cohesion at ages 3, 5, or 9 (98.73%). At each age (3, 5, and 9 years), we calculated z-scores for the violence exposure subscales (physical and emotional abuse, intimate partner violence, community violence exposure) and social deprivation subscales (physical and emotional neglect, romantic partner support, neighborhood social cohesion). The z-scores from each age were then averaged to create separate composite scores of violence exposure and social deprivation across ages 3, 5, and 9 (Hein, 2019). Confirmatory factor analysis revealed that our two factor model of violence exposure and social deprivation had good fit (χ2 = 655.94, df = 115, p < .05; CFI = 0.92; RMSEA = 0.03). Furthermore, the two factor model fit the data better than a one factor model that combined violence exposure and social deprivation (Satorra-Bentler scaled χ2 = 36.36, df = 1, p < .05) (Hein, 2019).

mental or neurological diagnoses (e.g., depression, attention-deficit/ hyperactivity disorder), and any recent medication use (e.g., Albuterol, Sertraline, hormonal birth control) were associated with the cortisol response to the SECPT. The FFCWS study design called for a 3:1 oversampling of non-marital births (Reichman et al., 2001). Therefore, we also tested whether mothers’ marital status at time of birth was associated with the cortisol response to the SECPT. Gender and age in months were reported by adolescents. For adolescents’ race/ethnicity, adolescents were binned according to whether they identified as Black or African American (n = 167), Caucasian/non-Hispanic (n = 29), or other (n = 26). Two dummy codes were created for adolescents’ race/ ethnicity, with youth identifying as Black or African American designated as the referent category in analyses. We used two dummy codes because the majority of adolescents in this sample identified as Black or African American and we expected the experiences of adolescents identifying as an ethnic or racial minority to differ from adolescents identifying as Caucasian. Time since waking was calculated as the difference in minutes between adolescents’ self-reported waking time and the time at which Sample 1 was collected. Baseline cortisol level was calculated as the average of cortisol at Samples 1 and 2. We included baseline cortisol as a covariate because each phase of the cortisol response to stress (reactivity slope, peak, recovery slope) is a function of starting levels of cortisol. Water temperature and seconds of hand immersion during the adolescents’ first attempt were recorded by an experimenter during the SECPT. Parents reported on adolescents’ mental or neurological diagnoses and medication use, which were recoded into dichotomous variables (“yes” or “no”). Birth mothers also reported their marital status when the child was born, which was recoded as either married (n = 48) or unmarried (n = 174). As a robustness check, mothers’ reports of annual household income from the FFCWS were included as a covariate to test whether childhood social deprivation was associated with adolescents’ cortisol response to the SECPT after controlling for household income during childhood. When adolescents were ages 3, 5, and 9, mothers reported their total household income in dollars from all sources in the last year. The FFCWS imputed dollar amounts for mothers with no reported income. At each age, less than 5% of income values were imputed for the FFCWS (SAND: n = 8 at age 3, n = 5 at age 5, n = 19 at age 9). Mother reports of household income at ages 3, 5, and 9 were then averaged and log transformed to account for positive skew of the data. 2.3. Statistical analysis Cortisol reactivity was modeled using two-piece growth curve modeling with landmark registration (GCM-LR) as applied to neuroendocrine data (Lopez-Duran et al., 2014). This approach accounts for individual differences in time to peak following a stressor and allows for examination of each phase of the cortisol response to stress (reactivity, peak, recovery). Specifically, we identified each adolescent’s peak cortisol concentration from their individual curves. Peaks were identified by selecting the first sample that was followed by a decline or plateau in cortisol concentration. If the first sample identified was followed by a plateau, the peak was the sample within the plateau with the largest cortisol concentration that was at least 10% above the first identified sample. A 15.5% increase in cortisol has been previously used as a meaningful cutoff to identify a cortisol response to social-evaluative stressors, including the SECPT (Miller et al., 2013; Schwabe and Schächinger, 2018). Therefore, if the adolescent’s peak represented at least a 15.5% increase from their baseline, the adolescent was identified as a responder and the time of the peak was used to create a new ‘peak time’ variable. For non-responders, the mode time of the peak of responders (25-min post-SECPT onset) was used to create the ‘peak time’ variable (Lopez-Duran et al., 2014). To account for positive skew, salivary cortisol concentrations were winsorized to the 98th percentile (n = 28 out of 1994 samples) and a Box-Cox power transformation for time series was applied to the data (Miller and Plessow, 2013). Two-

2.2.4. Covariates We tested whether adolescents’ gender, race/ethnicity, age in months, time since waking, baseline cortisol level, water temperature, seconds of hand immersion during their first attempt, presence of 4

Psychoneuroendocrinology 111 (2020) 104426

M.K. Peckins, et al.

piece multilevel growth curve models were fit with SAS® software (Version 9.4) using PROC MIXED (Singer, 1998) with restricted maximum likelihood estimation and an Autoregressive (1) covariance structure. An unconditional model was fit with the intercept set to each individual’s ‘peak time’ (non-responders set to the mode) and included baseline, minutes to peak (reactivity slope), and minutes after peak (recovery slope) with random intercepts and slopes to account for individual variability in these parameters (Lopez-Duran et al., 2014). Peak activation refers to the level of cortisol at the peak response of the curve. The reactivity and recovery slopes reflect the slope of cortisol toward and away from individuals’ peak activation, respectively (Lopez-Duran et al., 2014). We performed a series of Independent T-tests and chi-square tests to determine whether responders and non-responders significantly differed on the covariates gender, race/ethnicity, age, time since waking, water temperature, seconds of hand immersion during the adolescents’ first attempt, mental or neurological diagnoses, medication use, and mother’s marital status at time of child’s birth. We then tested whether the covariates were associated with the cortisol response to the SECPT in separate GCM-LR’s. All covariates were controlled for in subsequent analyses. Next, we performed GCM-LR’s to test whether violence exposure (with and without controlling for social deprivation) and social deprivation (with and without controlling for violence exposure) were associated with the cortisol response to the SECPT. Finally, we performed an exploratory analysis to test for an interaction between violence exposure and social deprivation. When there was a significant interaction, the Johnson-Neyman technique was used to determine the regions of significance using the Preacher et al. (2006) online computational tool.

Table 1 Descriptive characteristics of the Adolescent sample. Sample size

Gender Girls Boys Race/Ethnicity Black or African American, nonHispanic White or Caucasian, non-Hispanic Other Diagnosis of Mental or Neurological Condition Yes No Use of Asthma Medication Yes No Use of Birth control Yes No Use of Medication to Treat Depressive/ Anxiety Disorders Yes No Mother’s Marital Status at Birtha Married Cohabitating but not married Single Mother’s Marital Status at Age 9 Married Cohabitating but not married Single Average Household Incomeb Age (Months) Time Since Waking (Minutes) Hours of Sleep Water Temperature (Degrees) Time in Water During First Attempt (Seconds) Violence Exposure and Victimizationc Social Deprivationc Raw Cortisol Concentrations (μg/dL) Sample 1 (0 min pre-SECPT) Sample 2 (3 min post-SECPT) Sample 3 (15 min post-SECPT) Sample 4 (25 min post-SECPT) Sample 5 (30 min post-SECPT) Sample 6 (35 min post-SECPT) Sample 7 (40 min post-SECPT) Sample 8 (45 min post-SECPT) Sample 9 (60 min post-SECPT)

3. Results Descriptive information for key study variables are presented in Table 1. Descriptive information for the violence exposure and social deprivation subscales are presented in Supplementary Table S1. Pearson product-moment correlations for key study variables are presented in Supplementary Table S2. 3.1. Cortisol response to the SECPT On average, there was a 43.01% increase in cortisol concentrations from adolescents’ baseline to peak. There was an increase in cortisol by at least 15.5% from baseline to peak in 47.7% (n = 106) of the sample, and these adolescents were classified as responders (Fig. 1). Adolescents who did not increase in cortisol by at least 15.5% from baseline to peak were classified as non-responders (52.3%). The response rate reported in the present study is consistent with the response rate to the SECPT in adult samples (Schwabe and Schächinger, 2018). On average, responders and non-responders did not differ in gender, race/ethnicity, age, time since waking, water temperature, presence of mental or neurological diagnoses, medication use, and mothers’ marital status at time of child’s birth (ps = .10–.95). However, responder status was associated with seconds of hand immersion during the adolescents’ first attempt (t = 2.72, df = 198.70, p < .01) and mother reports of community violence exposure when adolescents were ages 3, 5, and 9 (χ2 = 3.99, df = 1, p < .05, V = .13). Adolescents classified as nonresponders kept their hand submerged in the ice water longer during their first attempt (M = 67.72 s, SD = 59.34) than responders (M = 49.30 s, SD = 39.76). Furthermore, 69.60% of non-responders’ mothers endorsed at least one form of community violence when adolescents were ages 3, 5, or 9 compared to 56.6% of responders. In models with only responders, there was a significant increase in cortisol from baseline to peak, b = .01, SE = .001, p < .001, and a significant post-peak decline in cortisol, b = −.01, SE = .001, p < .001. Peak activation was not significant (p = .90). In models with only non-responders, there was a significant decline in cortisol from baseline to

Mean (SD) or Frequency (%)

222 117 (52.7%) 105 (47.3%) 222 167 (75.23%) 29 (13.06%) 26 (11.71%) 221 33 (14.93%) 188 (85.07%) 222 22 (9.91%) 200 (90.09%) 222 7 (3.15%) 215 (96.85%) 222 9 (4.05%) 213 (95.95%) 222 48 (21.62%) 60 (27.03%) 114 (51.35%) 204

221 222 216 195 216 219

52 (23.42%) 17 (7.66%) 135 (60.81%) 35930.49 (33028.03) 189.90 (6.54) 406.44 (63.33) 7.51 (4.96) 33.57 (1.05) 58.88 (51.62)

222 222

0.005 (0.51) −0.001 (0.49)

220 222 222 222 221 222 222 222 221

0.15 0.14 0.15 0.16 0.16 0.14 0.13 0.12 0.12

(0.09) (0.09) (0.09) (0.11) (0.15) (0.10) (0.08) (0.09) (0.10)

Note. aMother’s marital status at birth was determined using data from the Fragile Families and Child Wellbeing Study (FFCWS) at birth (wave 1). b Average household income was calculated by taking the average of mother reports of household income at ages 3, 5, and 9 (waves 3, 4, and 5) using data from the FFCWS. cViolence exposure and victimization and social deprivation were composite scores calculated using data from the FFCWS at ages 3, 5, and 9 (waves 3, 4, and 5; Hein 2019).

peak, b = −.01, SE = .001, p < .001, and significant peak activation, b = −.43, SE = .03, p < .001. Non-responders’ post-peak decline in cortisol was not statistically significant (p = .05). In models including both responders and non-responders, there was no significant change in cortisol from baseline to peak; however, there was a significant postpeak decline in cortisol. The lack of significant change from baseline to peak is being driven by the flat cortisol response of non-responders. All subsequent models include both responders and non-responders. Gender, time since waking, water temperature, and mothers’ marital status at child’s time of birth were not associated with the cortisol response to the SECPT (ps = .13–.99). Adolescents identifying as Black or African American had increased cortisol concentrations at baseline, b = 5

Psychoneuroendocrinology 111 (2020) 104426

M.K. Peckins, et al.

Fig. 1. Cortisol response to the Socially Evaluated Cold-Pressor Task (SECPT; Schwabe et al., 2008) for Responders (n = 106) and NonResponders (n = 116). Responders had at least a 15.5% increase in cortisol concentrations from their baseline to peak following the SECPT. Cortisol concentrations were winsorized to the 98th percentile.

reactivity slope, b = -0.01, SE = 0.001, p < .001, decreased peak activation, b = -0.28, SE = 0.07, p < .001, and less steep recovery slope, b = 0.003, SE = 0.001, p < .01. Violence exposure was associated with a blunted cortisol response to the SECPT even when controlling for the effects of social deprivation and all covariates.

0.19, SE = 0.08, p < .05, and at peak activation, b = 0.40, SE = 0.14, p < .01, compared to adolescents identifying as any other race or ethnicity. Age in months was negatively associated with cortisol concentrations at baseline, b = -0.01, SE = 0.004, p < .01, and at peak activation, b = -0.02, SE = 0.01, p < .05. Presence of a mental or neurological disorder (bbaseline = 0.22, SE = 0.07, p < .01; bpeak = 0.28, SE = 0.12, p < .05) and medication use (bbaseline = 0.11, SE = 0.05, p < .05; bpeak = 0.25, SE = 0.09, p < .01) were both associated with increased cortisol concentrations at baseline and peak activation, respectively. Longer time of hand immersion during adolescents’ first attempt at the SECPT was associated with a less steep cortisol reactivity slope, b = -5.0 × 10−5, SE = 1.4 × 10−5, p < .001.

3.3. Social deprivation Greater levels of social deprivation experienced during childhood were associated with decreased peak activation, b = -0.32, SE = 0.09, p < .001 when controlling for all covariates (Table 2). Social deprivation was not associated with cortisol reactivity (p = .17) and cortisol recovery (p = .59). As a robustness check, we confirmed that social deprivation was still associated with decreased peak activation (p < .01) after controlling for mothers’ mean household income during childhood. Social deprivation was no longer associated with peak activation when controlling for violence exposure during childhood. This indicates that the association between social deprivation and peak activation may be due to co-occurring violence exposure.

3.2. Violence exposure Greater levels of violence exposure experienced during childhood were associated with an overall blunted cortisol response to the SECPT in adolescence when controlling for all covariates (Table 2). Specifically, increased violence exposure was associated with a less steep

Table 2 Estimates for adjusted growth curve models of the cortisol response to stress predicted by violence exposure and social deprivation. Dimension of Childhood Adversity Violence Exposure

Intercept Baseline Cortisol Time Before Peak Time After Peak Violence Exposure Baseline Cortisol x Violence Exposure Time Before Peak x Violence Exposure Time After Peak x Violence Exposure Social Deprivation Baseline Cortisol x Social Deprivation Time Before Peak x Social Deprivation Time After Peak x Social Deprivation

Social Deprivation

Violence Exposure and Social Deprivation

b

SE

t-value

b

SE

t-value

b

SE

t-value

−0.23 0.81 0.005 −0.01 −0.28 −0.10 −0.01 0.003 — — — —

0.12 0.07 0.002 0.002 0.07 0.05 0.001 0.001 — — — —

−1.99* 11.58** 2.30* −3.30** −3.73** −2.23* −3.65** 2.60** — — — —

−0.28 0.78 0.004 −0.01 — — — — −0.32 −0.18 −0.002 −0.001

0.12 0.07 0.002 0.002 — — — — 0.09 0.06 0.001 0.001

−2.34* 10.87** 2.17* −3.29** — — — — −3.41** −3.14** −1.37 −0.54

−0.28 0.78 0.005 −0.01 −0.21 −0.06 −0.01 0.004 −0.19 −0.13 8.80 × 10−5 −0.002

0.12 0.07 0.002 0.002 0.08 0.05 0.002 0.001 0.11 0.07 0.002 0.001

−2.32* 10.93** 2.31* −2.41** −2.45* −1.08 −3.38** 3.04** −1.77 −1.96 0.06 −1.68

Note. * p < .05, ** p < .01. Intercept reflects peak activation, Time Before Peak represents cortisol reactivity slope, and Time After Peak represents cortisol recovery slope. Baseline Cortisol is the average of cortisol concentrations at Samples 1 and 2. Violence Exposure and Social Deprivation were composite scores calculated using data from FFCWS at waves 3, 4, and 5 (Hein, 2019). Models control for gender, race/ethnicity, age, time since waking, water temperature, seconds of hand immersion during the adolescents’ first attempt, presence of mental or neurological diagnoses, medication use, and mothers’ marital status at time of child’s birth. 6

Psychoneuroendocrinology 111 (2020) 104426

M.K. Peckins, et al.

Table 3 Estimates for adjusted growth curve models of the cortisol response to stress predicted by the interaction between violence exposure and social deprivation.

Intercept Baseline Cortisol Time Before Peak Time After Peak Violence Exposure and Victimization Baseline Cortisol x Violence Exposure Time Before Peak x Violence Exposure Time After Peak x Violence Exposure Social Deprivation Baseline Cortisol x Social Deprivation Time Before Peak x Social Deprivation Time After Peak x Social Deprivation Violence Exposure x Social Deprivation Baseline Cortisol x Violence Exposure x Social Deprivation Time Before Peak x Violence x Social Deprivation Time After Peak x Violence x Social Deprivation

b

SE

t-value

−0.32 0.76 0.004 −0.01 −0.30 −0.10 −0.01 0.004 −0.27 −0.17 −0.001 −0.002 0.47 0.24

0.12 0.07 0.002 0.002 0.09 0.06 0.002 0.001 0.11 0.07 0.002 0.002 0.21 0.13

−2.69** 10.68** 2.08* −3.38** −3.30** −1.84 −3.65** 2.98** −2.37* −2.35* −0.47 −1.59 2.21* 1.83

0.003

0.002

1.53

−1.30 × 10-4

0.002

0.94

Fig. 2. 95% confidence bands for observed sample values of social deprivation. At lower levels of social deprivation, violence exposure has a blunting effect on peak activation of the HPA-axis (shaded area to the left of the dashed line). At higher levels of social deprivation, violence exposure is no longer associated with peak activation of the HPA-axis (area to the right of the dashed line).

Note. * p < .05, ** p < .01. Intercept reflects peak activation, Time Before Peak represents cortisol reactivity slope, and Time After Peak represents cortisol recovery slope. Baseline Cortisol is the average of cortisol concentrations at Samples 1 and 2. Violence Exposure and Social Deprivation were composite scores calculated using data from FFCWS at waves 3, 4, and 5 (Hein, 2019). Models control for gender, race/ethnicity, age, time since waking, water temperature, seconds of hand immersion during the adolescents’ first attempt, presence of mental or neurological diagnoses, medication use, and mothers’ marital status at time of child’s birth.

3.4. Interaction between violence exposure and social deprivation There was a significant two-way interaction between childhood violence exposure and social deprivation predicting peak activation, b = 0.47, SE = 0.21, p < .05, when controlling for all covariates (Table 3). Increased social deprivation during childhood was associated with a reduced blunting effect of violence exposure on peak activation, and the effect of violence exposure on peak activation was no longer statistically significant at levels of social deprivation approximately 1 SD above the mean (Figs. 2 and 3). Approximately 19% of the sample had levels of social deprivation within this region of significance. There was no moderating effect of social deprivation on the association between violence exposure and reactivity slope (p = .13) and recovery slope (p = .94). 4. Discussion

Fig. 3. Significant interaction between violence exposure and social deprivation predicting peak activation of the HPA-axis. At low (1 SD below the mean; b = -0.42, p < .05) and mean levels of social deprivation (b = -0.30, p < .01), greater levels of violence exposure are associated with lower peak activation of the HPA-axis. At high levels of social deprivation (1 SD above the mean), violence exposure is not associated with peak activation of the HPA-axis (b = -0.19, p = .08).

The present study elucidated how two dimensions of childhood adversity, measured prospectively, impacted the HPA-axis response to stress in a diverse birth-cohort study of adolescents, many of whom have faced substantial adversity. We found that violence exposure experienced during childhood was associated with overall blunting of the HPA-axis, above and beyond the effects of social deprivation and a myriad of relevant covariates. Social deprivation was associated with peak cortisol activation in univariate models, but was not associated with the cortisol response to the SECPT when controlling for experiences of violence. However, the blunting effect of violence exposure on peak HPA-axis activation was conditional on social deprivation. Although exploratory, this finding suggests that higher levels of social deprivation diminish the blunting effect of violence exposure on peak HPA-axis activation. These findings can inform the study of socioeconomically disadvantaged samples of youth and provide insight into the ways different dimensions of childhood adversity impact HPA-axis reactivity. The blunting effect of violence exposure on the HPA-axis response to

stress may be an adaptive response to chronic activation of the HPAaxis (Susman, 2006). For example, short-term elevations in cortisol modulate the immune response, reproductive and growth axes, and mobilization of energy; yet prolonged exposure to cortisol can increase risk for health and behavior problems across the lifespan (Agorastos et al., 2018; McEwen, 1998). Thus, a downregulated HPA-axis may be protective in the context of chronic stress, such as violence exposure (Del Giudice et al., 2011; Fries et al., 2005). Although we were not able to test the mechanisms, downregulation of the HPA-axis may be due to increased negative feedback sensitivity (Fries et al., 2005; Heim et al., 2000). Our findings support the DMAP (McLaughlin et al., 2014) and expand on previous tests of DMAP (Busso et al., 2017) by 7

Psychoneuroendocrinology 111 (2020) 104426

M.K. Peckins, et al.

ambiguity as threatening, contributing to increased physiological reactivity to ambiguous situations (Chen et al., 2004). Thus, given the high-risk nature of our sample, non-response to the SECPT may reflect blunting of the HPA-axis or differences in adolescents’ cognitive appraisal of the SECPT as threatening. The present study has many strengths, including a relatively large, socioeconomically diverse sample of youth with many adolescents from groups under-represented in this area of research (i.e., African Americans); the use of prospective, dimensional longitudinal data related to early life experiences; the collection of 9 saliva samples throughout the SECPT; and the use of two-piece GCM-LR to account for individual variability in the time at which adolescents’ reached peak cortisol concentrations. However, results from the present study should be considered within the context of some limitations. First, research assistants who observed the SECPT were predominantly Caucasian and female, so we do not know how race and gender may have influenced participants’ perceptions of social evaluation and cortisol response to the SECPT. Second, our measures of violence exposure and social deprivation were based on prospective reports from parents because children from the FFCWS were not asked to report on their experiences before the age of 9. Therefore, our measures of violence exposure and social deprivation may not capture all of the child’s experiences in other settings outside of the home. Finally, we do not have the same measures of violence exposure and social deprivation when children were between the ages of 9 and 15. Thus, we were not able to test how our constructs of violence exposure and social deprivation during late childhood and early adolescence impact the HPA-axis response to stress. Despite these limitations, results from the present study provide valuable information on how childhood experiences may influence HPA-axis reactivity to the SECPT during adolescence: Few studies have tested how different types of adversity during childhood impact each phase of the HPA-axis response to stress in the same model. Moreover, few studies have examined these questions in adolescents in the community, particularly via a well-sampled cohort enriched for exposure to adversity and with substantial representation of African American adolescents. Therefore, findings from the present study have several important implications. First, our study demonstrated that childhood experiences high in violence exposure, but not social deprivation, were associated with blunting of the HPA-axis response to stress in adolescence, which is not consistent with existing research on the negative effects of poverty and child neglect. Second, the blunting effect of violence exposure on peak activation diminished in the context of high social deprivation. These findings highlight the importance of considering both dimensions when studying the consequences of adversity for HPA-axis functioning. Future studies should continue testing how different dimensions of adversity uniquely predict HPA-axis functioning in diverse samples of adolescents enriched for socioeconomic disadvantage, and what impact they may have on risk for health and behavior problems during adolescence and adulthood. Future studies should also consider how the timing and chronicity of different dimensions of childhood adversity impact HPA-axis functioning during adolescence.

demonstrating that violence has a blunting effect on each phase of the HPA-axis response to stress. Our findings conflict with previous research on social and environmental deprivation that has found substantial evidence for both hyper- and hypo-activation of the HPA-axis in response to stress following these experiences (reviewed in Koss and Gunnar, 2018). However, experiences of social and environmental deprivation are also confounded with increased risk for experiences high in threat (Coulton et al., 2007), such as interpersonal violence and physical abuse, which have rarely been addressed in previous studies. Interestingly, we found a link between social deprivation and blunting of the HPA-axis that was due to the high levels of violence experienced by youth. That is, had we not assessed both types of experiences, we might have concluded that social deprivation was related to HPA-axis functioning. Indeed, violence exposure is likely a more reliable activator of the HPA-axis than other types of deprivation (e.g., lack of social stimulation). However, some aspects of deprivation, such as hunger, may lead to increased activation of the HPA-axis and thus later blunting (Tomiyama et al., 2010). This may explain why children with severe forms of deprivation (e.g., institutionalized rearing; Gunnar et al., 2001) show dysregulated HPA-axis functioning while other studies, including the present one, do not find such a link using a more broad definition of deprivation. Rather, the present study demonstrated that the effects of social deprivation on the HPA-axis may actually be due to confounding violence exposure. Interestingly, the blunting effect of violence exposure on peak activation was no longer present at high levels of social deprivation. Although this analysis was exploratory in nature and should be interpreted with caution, this result fits within the framework of the Adaptive Calibration Model (Del Giudice et al., 2011), which suggests a buffered response (i.e., moderate blunting) is adaptive in the context of moderate levels of stress. In more dangerous and unpredictable environments, a vigilant response (i.e., elevated reactivity) is more adaptive. In the present study, the combination of both violence exposure and social deprivation may signal a more dangerous and unpredictable environment and thus, blunting may no longer be adaptive. This interpretation is consistent with other studies that have found higher levels of negative family relationships and lower levels of familial warmth and predictability to be more strongly associated with a vigilant physiological response than a buffered physiological response (Del Giudice et al., 2012). A noteworthy finding is that the SECPT elicited at least a 15.5% increase in cortisol in 47.7% of our sample, whereas the SECPT validation study found a 40% increase in cortisol from baseline levels in 70% of their sample (Schwabe et al., 2008). Previous studies have reported a response rate of 40% to the CPT without a social-evaluative component (Schwabe et al., 2008; Smeets et al., 2012), suggesting a subset of our sample may not have perceived the video camera and research assistant observing the SECPT as socially-evaluative. This is important as a meta-analysis found that laboratory stressors perceived as uncontrollable or characterized by social-evaluative threat were most effective at eliciting a cortisol response (Dickerson and Kemeny, 2004). Our sample identified predominantly as African American, and there is some evidence that suggests African Americans have a reduced response to the Trier Social Stress Test (TSST; Kirschbaum et al., 1993) that may be attributed to biological and genetic differences (Chong et al., 2008) or race-related stressors (Richman and Jonassaint, 2008). Unfortunately, we did not ask adolescents to report their level of stress after completing the SECPT, so we are not able to test whether adolescents were less stressed by the task that affected their cortisol response to the SECPT. A second explanation for the low response rate is that the present study included adolescents who were oversampled based on their parents’ marital status which is highly correlated with socioeconomic hardship (Reichman et al., 2001). Prior research has shown adolescents from low socioeconomic status backgrounds are more likely to interpret

Declaration of Competing Interest The authors declare that there are no conflicts of interest. Acknowledgments The research reported in this paper was supported by grants from the National Institutes of Health, R01MH103761(Monk) and T32HD007109(McLoyd and Monk), and a Doris Duke Fellowship for the Promotion of Child Well-Being (Hein). We would also like to acknowledge the past work of the Fragile Families and Child Wellbeing Study, the families for sharing their experiences with us, and the project 8

Psychoneuroendocrinology 111 (2020) 104426

M.K. Peckins, et al.

staff for making this study possible.

tool for investigating psychobiological stress responses in a laboratory setting. Neuropsychobiology 28, 76–81. https://doi.org/10.1159/000119004. Koss, K.J., Gunnar, M.R., 2018. Annual research review: early adversity, the hypothalamic–pituitary–adrenocortical axis, and child psychopathology. J. Child Psychol. Psychiatry 59, 327–346. https://doi.org/10.1111/jcpp.12784. Kuhlman, K.R., Geiss, E.G., Vargas, I., Lopez-Duran, N.L., 2015. Differential associations between childhood trauma subtypes and adolescent HPA-axis functioning. Psychoneuroendocrinology 54, 103–114 https://dx.doi.org/ 10.1016%2Fj.psyneuen.2015.01.020. Lopez-Duran, N.L., Mayer, S.E., Abelson, J.L., 2014. Modeling neuroendocrine stress reactivity in salivary cortisol: adjusting for peak latency variability. Stress 17, 285–295. https://doi.org/10.3109/10253890.2014.915517. Manuel, J.I., Martinson, M.L., Bledsoe-Mansori, S.E., Bellamy, J.L., 2012. The influence of stress and social support on depressive symptoms in mothers with young children. Soc. Sci. Med. 75, 2013–2020. https://doi.org/10.1016/j.socscimed.2012.07.034. McEwen, B.S., 1998. Stress, adaptation, and disease: allostasis and allostatic load. Ann. N. Y. Acad. Sci. 840, 33–44. https://doi.org/10.1111/j.1749-6632.1998.tb09546.x. McLaughlin, K.A., Sheridan, M.A., Lambert, H.K., 2014. Childhood adversity and neural development: deprivation and threat as distinct dimensions of early experience. Neurosci. Biobehav. Rev. 47, 578–591 https://dx.doi.org/ 10.1016%2Fj.neubiorev.2014.10.012. McLaughlin, K.A., Sheridan, M.A., Nelson, C.A., 2017. Neglect as a violation of speciesexpectant experience: neurodevelopmental consequences. Biol. Psychiatry 82, 462–471 https://dx.doi.org/10.1016%2Fj.biopsych.2017.02.1096. Miller, R., Plessow, F., 2013. Transformation techniques for cross-sectional and longitudinal endocrine data: application to salivary cortisol concentrations. Psychoneuroendocrinology 38, 941–946. https://doi.org/10.1016/j.psyneuen.2012. 09.013. Miller, R., Plessow, F., Kirschbaum, C., Stalder, T., 2013. Classification criteria for distinguishing cortisol responders from nonresponders to psychosocial stress: evaluation of salivary cortisol pulse detection in panel designs. Psychosom. Med. 75, 832–840. https://doi.org/10.1097/PSY.0000000000000002. Morenoff, J.D., Sampson, R.J., Raudenbush, S.W., 2001. Neighborhood inequality, collective efficacy, and the spatial dynamics of urban violence. Criminology 39, 517–558. https://doi.org/10.1111/j.1745-9125.2001.tb00932.x. Preacher, K.J., Curran, P.J., Bauer, D.J., 2006. Computational tools for probing interactions in multiple linear regression, multilevel modeling, and latent curve analysis. J. Educ. Behav. Stat. 31, 437–448 https://doi.org/10.3102%2F10769986031004437. Reichman, N.E., Teitler, J.O., Garfinkel, I., McLanahan, S.S., 2001. Fragile families: sample and design. Child. Youth Serv. Rev. 23, 303–326. https://doi.org/10.1016/ S0190-7409(01)00141-4. Richman, L.S., Jonassaint, C., 2008. The effects of race-related stress on cortisol reactivity in the laboratory: implications of the Duke lacrosse scandal. Ann. Behav. Med. 35, 105–110. https://doi.org/10.1007/s12160-007-9013-8. Roberts, A.G., Lopez-Duran, N.L., 2019. Developmental influences on stress response systems: implications for psychopathology vulnerability in adolescence. Compr. Psychiatry 88, 9–21. https://doi.org/10.1016/j.comppsych.2018.10.008. Schwabe, L., Haddad, L., Schachinger, H., 2008. HPA axis activation by a socially evaluated cold-pressor test. Psychoneuroendocrinology 33, 890–895. https://doi.org/10. 1016/j.psyneuen.2008.03.001. Schwabe, L., Schächinger, H., 2018. Ten years of research with the socially evaluated cold pressor test: data from the past and guidelines for the future. Psychoneuroendocrinology 92, 155–161. https://doi.org/10.1016/j.psyneuen.2018. 03.010. Singer, J.D., 1998. Using SAS PROC MIXED to fit multilevel models, hierarchical models, and individual growth models. J. Educ. Behav. Stat. 23, 323–355 https://doi.org/ 10.3102%2F10769986023004323. Smeets, T., Cornelisse, S., Quaedflieg, C.W.E.M., Meyer, T., Jelicic, M., Merckelbach, H., 2012. Introducing the Maastricht Acute Stress Test (MAST): a quick and non-invasive approach to elicit robust autonomic and glucocorticoid stress responses. Psychoneuroendocrinology 37, 1998–2008. https://doi.org/10.1016/j.psyneuen. 2012.04.012. Straus, M.A., Hamby, S.L., Finkelhor, D., Moore, D.W., Runyan, D., 1998. Identification of child maltreatment with the Parent-Child Conflict Tactics Scales: development and psychometric data for a national sample of American parents. Child Abuse Negl. 22, 249–270. https://doi.org/10.1016/S0145-2134(97)00174-9. Susman, E.J., 2006. Psychobiology of persistent antisocial behavior: stress, early vulnerabilities and the attenuation hypothesis. Neurosci. Biobehav. Rev. 30, 376–389. https://doi.org/10.1016/j.neubiorev.2005.08.002. Tomiyama, A.J., Mann, T., Vinas, D., Hunger, J.M., Dejager, J., Taylor, S.E., 2010. Low calorie dieting increases cortisol. Psychosom. Med. 72, 357–364 https://dx.doi.org/ 10.1097%2FPSY.0b013e3181d9523c. Trickett, P.K., Noll, J.G., Susman, E.J., Shenk, C.E., Putnam, F.W., 2010. Attenuation of cortisol across development for victims of sexual abuse. Dev. Psychopathol. 22, 165–175. https://doi.org/10.1017/S0954579409990332. Wismer Fries, A.B., Shirtcliff, E.A., Pollak, S.D., 2008. Neuroendocrine dysregulation following early social deprivation in children. Dev. Psychobiol. 50, 588–599. https:// doi.org/10.1002/dev.20319. Zhang, S., Anderson, S.G., 2010. Low-income single mothers’ community violence exposure and aggressive parenting practices. Child. Youth Serv. Rev. 32, 889–895. https://doi.org/10.1016/j.childyouth.2010.02.010.

Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.psyneuen.2019. 104426. References Agorastos, A., Pervanidou, P., Chrousos, G.P., Kolaitis, G., 2018. Early life stress and trauma: developmental neuroendocrine aspects of prolonged stress system dysregulation. Hormones 17, 507–520. https://doi.org/10.1007/s42000-018-0065-x. Aiyer, S.M., Heinze, J.E., Miller, A.L., Stoddard, S.A., Zimmerman, M.A., 2014. Exposure to violence predicting cortisol response during adolescence and early adulthood: understanding moderating factors. J. Youth Adolesc. 43, 1066–1079. https://doi.org/ 10.1007/s10964-014-0097-8. Barrington, W.E., Stafford, M., Hamer, M., Beresford, S.A.A., Koepsell, T., Steptoe, A., 2014. Socioeconomic deprivation, perceived neighborhood factors, and cortisol responses to induced stress among healthy adults. Health Place 27, 120–126. https:// doi.org/10.1016/j.healthplace.2014.02.001. Bosch, N.M., Riese, H., Reijneveld, S.A., Bakker, M.P., Verhulst, F.C., Ormel, J., Oldehinkel, A.J., 2012. Timing matters: long term effects of adversities from prenatal period up to adolescence on adolescents’ cortisol stress response. The TRAILS study. Psychoneuroendocrinology 37, 1439–1447. https://doi.org/10.1016/j.psyneuen. 2012.01.013. Busso, D.S., McLaughlin, K.A., Sheridan, M.A., 2017. Dimensions of adversity, physiological reactivity, and externalizing psychopathology in adolescence: deprivation and threat. Psychosom. Med. 79, 162–171 https://dx.doi.org/ 10.1097%2FPSY.0000000000000369. Chen, E., Langer, D.A., Raphaelson, Y.E., Matthews, K.A., 2004. Socioeconomic status and health in adolescents: the role of stress interpretations. Child Dev. 75, 1039–1052. https://doi.org/10.1111/j.1467-8624.2004.00724.x. Chong, R.Y., Uhart, M., McCaul, M.E., Johnson, E., Wand, G.S., 2008. Whites have a more robust hypothalamic–pituitary–adrenal axis response to a psychological stressor than blacks. Psychoneuroendocrinology 33, 246–254. https://doi.org/10.1016/j. psyneuen.2007.10.014. Coulton, C.J., Crampton, D.S., Irwin, M., Spilsbury, J.C., Korbin, J.E., 2007. How neighborhoods influence child maltreatment: a review of the literature and alternative pathways. Child Abuse Negl. 31, 1117–1142. https://doi.org/10.1016/j. chiabu.2007.03.023. Del Giudice, M., Ellis, B.J., Shirtcliff, E.A., 2011. The adaptive calibration model of stress responsivity. Neurosci. Biobehav. Rev. 35, 1562–1592. https://doi.org/10.1016/j. neubiorev.2010.11.007. Del Giudice, M., Hinnant, J.B., Ellis, B.J., El-Sheikh, M., 2012. Adaptive patterns of stress responsivity: a preliminary investigation. Dev. Psychol. 48, 775–790 https:// dx.doi.org/10.1037%2Fa0026519. Dickerson, S.S., Kemeny, M.E., 2004. Acute stressors and cortisol responses: a theoretical integration and synthesis of laboratory research. Psychol. Bull. 130, 355–391. Fries, E., Hesse, J., Hellhammer, J., Hellhammer, D.H., 2005. A new view on hypocortisolism. Psychoneuroendocrinology 30, 1010–1016. https://doi.org/10.1016/j. psyneuen.2005.04.006. Goetschius, L.G., Hein, T.C., Mattson, W.I., Lopez-Duran, N., Dotterer, H.L., Welsh, R.C., Mitchell, C., Hyde, L.W., Monk, C.S., 2019. Amygdala-prefrontal cortex white matter tracts are widespread, variable and implicated in amygdala modulation in adolescents. NeuroImage 191, 278–291. https://doi.org/10.1016/j.neuroimage.2019.02. 009. Heim, C., Ehlert, U., Hellhammer, D.H., 2000. The potential role of hypocortisolism in the pathophysiology of stress-related bodily disorders. Psychoneuroendocrinology 25, 1–35. https://doi.org/10.1016/S0306-4530(99)00035-9. Hein, T.C., Mattson, W.I., Dotterer, H.L., Mitchell, C., Lopez-Duran, N., Thomason, M.E., Peltier, S.J., Welsh, R.C., Hyde, L.W., Monk, C.S., 2018. Amygdala habituation and uncinate fasciculus connectivity in adolescence: a multi-modal approach. NeuroImage 183, 617–626. https://doi.org/10.1016/j.neuroimage.2018.08.058. Hein, T.C., 2019. Dimensions of early adversity as distinct predictors of adolescent brain development (Doctoral Dissertation, University of Michigan). Retrieved from: https://deepblue.lib.umich.edu/bitstream/handle/2027.42/149945/heint_1.pdf? sequence=1&isAllowed=y. Hunt, T.K.A., Slack, K.S., Berger, L.M., 2017. Adverse childhood experiences and behavioral problems in middle childhood. Child Abuse Negl. 67, 391–402 https:// dx.doi.org/10.1016%2Fj.chiabu.2016.11.005. James, S., Donnelly, L., Brooks-Gunn, J., McLanahan, S., 2018. Links between childhood exposure to violent contexts and risky adolescent health behaviors. J. Adolesc. Health 63, 94–101. https://doi.org/10.1016/j.jadohealth.2018.01.013. Kapuku, G.K., Treiber, F.A., Davis, H.C., 2002. Relationships among socioeconomic status, stress induced changes in cortisol, and blood pressure in African American males. Ann. Behav. Med. 24, 320–325. https://doi.org/10.1207/ S15324796ABM2404_08. Kirschbaum, C., Pirke, K.M., Hellhammer, D.H., 1993. The ‘Trier Social Stress Test’ – a

9